Researchers evaluate deep-learning and face recognition

A team of scientists from Istanbul, Turkey have completed an evaluation of deep-learning based approaches, stating that these have been dominating the face recognition field due to the significant performance improvement they have provided on the challenging wild datasets.Mostafa Mehdipour Ghazia and Hazım Kemal Ekenel, from the Sabanci University, and Istanbul Technical University have evaluated the performance of deep learning based face representation under several conditions, including the varying head pose angles, upper and lower face occlusion, changing illumination of different strengths, and misalignment due to erroneous facial feature localization.Prior to an emergence of deep learning algorithms, the majority of traditional face recognition methods used to first locally extract hand-crafted shallow features from facial images using Local Binary Patterns (LBP), Scale Invariant Feature Transform (SIFT), and Histogram of Oriented Gradients (HOG), note Ghazia and Ekenel.The researchers used two successful and publicly available deep learning models: VGG-Face and Lightened CNN.While the pair noted that the obtained results showed that although deep learning provides a powerful representation for face recognition, it can still benefit from pre-processing, for example, for pose and illumination normalisation for better performance.”We have found that deep learning based face representation is robust to misalignment and able to tolerate facial feature localization errors up to 10% of the interocular distance. … The VGG-Face model is shown to be more transferable compared to the Lightened CNN model”Overall, we believe that deep learning based face recognition requires further research to address the problem of face recognition under mismatched conditions, especially when there is a limited amount of data available for the task at hand.”

Also in the news:

ImageWare Systems announces new anti-spoofing technology Report shows Facebook lobbying growth ViewSonic gives glimpse of smartphone with iris scanner